betaFeedback

WorksPhysics and Astronomy

Electron-phonon coupling in magnetic materials using the local spin density approximation

Abstract

Magnetic materials are crucial for manipulating electron spin and magnetic fields, enabling applications in data storage, spintronics, charge transport, and energy conversion, while also providing insight into fundamental quantum phenomena. In numerous applications, the interaction between electrons and lattice vibrations, known as electron-phonon coupling, can be of significant importance. In that regard, we extend the EPW package to be able to interpolate the electron-phonon matrix elements combining perturbation theory and maximally localized Wannier functions. This allows to use dense momentum grids at a reasonable computational cost when computing electron-phonon-related quantities and physical properties. We validate our implementation considering ferromagnetic iron and nickel, where we explore the phonon induced mass enhancement and Eliashberg spectral function finding different importance of each spin channel for both compounds. Furthermore, we evaluate the carrier resistivity at finite temperatures for both systems, considering the role of the magnetic phase in carrier transport. Our findings indicate that in the case of Fe, the primary contributor to resistivity is electron-phonon scattering. In contrast, for Ni, electron-phonon scattering constitutes less than one-third of the resistivity, underscoring a fundamental difference in the transport properties of the two systems.

manuscript.pdf

Electron-phonon coupling in magnetic materials using the local spin density approximation

Introduction

Magnetic materials play an important role for technological applications, including magnetic cores in electric motors and generators, or magnetic memory devices in computers and mobile phones. From a historical standpoint, iron is arguably the most significant magnetic material. Iron can be alloyed or combined with other elements to form compounds that are metals, insulators [1], superconductors [2], strong ferromagnetism [3], antiferromagnetism [4], thermal shielding at low and high temperatures and has good mechanical properties [5].

Interest in magnetic materials is at an all-time high due to the end of Moore’s law [6] and other limitations that current electronic devices are facing. In that regard, magnetic materials are seen as an appealing solution leveraging the spin of the electron giving rise to the field of spintronics. Spintronic devices exploit both the electronic and spin degrees of freedom [7, 8, 9], offering promising logic and memory architectures with inherently low power consumption [10, 11, 12, 13]. In addition, spintronic devices are widely used for memory applications [14, 15, 16, 17]. Iron and nickel are ubiquitous in steels, alloys, and functional magnetic devices used in the energy sector. Thus, there is renewed effort to understand, search and characterize potential materials relevant for the mentioned applications [18, 19, 20, 21]. Apart from these advances in devices for classical computers, the field of quantum computing might also require magnetic compounds that present specific transport properties [22, 23], and a certain robustness to thermal noise to build and shield the quantum bits [24] [25].

Given that electrical and thermal conductivity in metals and semiconductors is strongly influenced by electron-phonon scattering, developing accurate first-principles methods to capture this mechanism is central to understanding and reducing energy losses in conductive materials. Although electron-phonon interactions have been studied for decades, accurately describing them has proven to be a challenging task. The situation becomes even more complicated when one considers the effect that phonons can have on the local magnetic structure [26–30], where in some cases they can even induce a non-zero transient magnetization in non-magnetic materials [31, 32]. As a result, it is common to rely on effective models that take input from experiments [33–35], or previously computed ab initio quantities [36–39]. Today, first-principles electron-phonon interactions are widely available thanks to density functional theory (DFT) [40, 41] and density functional perturbation theory (DFPT) [42–45]. However, this approach is not free from challenges that can make actual calculations impractical. One of the main bottlenecks is the high computational cost required to sample the Brillouin zone in both electronic kk and phononic qq momentum space when integrating the different observables. In this regard, one of the most widespread solutions is to combine DFPT with maximally localized Wannier functions [46, 47], which enables interpolation at arbitrary momentum k\mathbf{k} and q\mathbf{q} of the electron-phonon matrix elements (EPME) gmnν(k,q)g_{mn\nu}(\mathbf{k}, \mathbf{q}), electronic eigenvalues εnk\varepsilon_{n\mathbf{k}} and eigenvectors ψnk\psi_{n\mathbf{k}}. Most implementations to date assume time-reversal symmetry. This assumption is only suited to the study of non-magnetic materials, hindering a global understanding of electron-phonon interactions. In fact, this is an important barrier to understanding the role of phonons in spintronics, caloritronics, or superconductivity and charge density waves in the presence of magnetism. More importantly, both electron and phonon dispersions can be affected by including the spin degree of freedom in the calculations while retaining the same exchange and correlation functional [48]. For example, neglecting the spin of iron leads to unstable phonons [49].

This limitation has been addressed using finite differences [48, 50–57] but is limited due to the computational cost of supercells. Recently, support for DFPT with non-collinear magnetism and ultrasoft pseudopotentials has been implemented in Abinit [49, 58–61] and Quantum ESPRESSO (QE) [62–64], as well as interpolation of the electron-phonon matrix elements with collinear magnetism in Perturbo [65–68]. We here present a first-principles implementation for electron–phonon properties in magnetic systems, validated and applied to Fe and Ni. Our results reveal the essential role of the magnetic phase in resistivity, the microscopic origin of scattering in different ferromagnets, and the absence of phonon-mediated superconductivity in both materials. These insights provide a foundation for designing magnetic materials with tailored transport properties for energy applications. In this work, we extend the capabilities of the electron phonon Wannier (EPW) code [46, 69, 70] to support collinear magnetism, unlocking the path to study electron-phonon related properties in magnetic materials with high accuracy and at reasonable computational cost. To this end, we exploit recent developments in the Abinit code on spin-dependent electron-phonon interactions [59] to validate our implementation. This choice is guided by the documented small discrepancy between Abinit and Quantum ESPRESSO [71–73]. Although collinear magnetism serves as an approximation to its non-collinear counterpart when considering spin-orbit coupling (SOC), its computational efficiency is markedly superior, while retaining a good degree of accuracy.

We study the carrier transport properties of ferromagnetic (FM) body-centered cubic (BCC) Fe and face-centered cubic (FCC) Ni when electron phonon interactions are accounted. Our findings indicate that the omission of magnetism in Fe leads to the emergence of imaginary (soft) phonon modes, confirming previous works. As sometimes done in the literature [55, 74–76], excluding imaginary phonon modes overestimated resistivity due to the large contribution of modes on the onset of being imaginary. In contrast, Ni is a weaker ferromagnet with similar ferromagnetic and non-magnetic phonon band dispersions, but with a very different change of resistivity with temperature. Overall, the inclusion of magnetism reduces resistivity by a factor of three and seven for Fe and Ni, respectively. Finally, we compute the phonon-driven superconducting transition temperature (TCT_{\mathrm{C}}) in Fe and Ni and find vanishingly small values, which shows that neither material would be superconducting, even if their Cooper pairs could survive in the FM magnetization field.

This work opens the gate towards accounting for the role of phonons in magnetic systems and possible superconducting states in magnetic materials (such as iron arsenides and possibly cuprates), which are highly relevant for new energy-efficient technologies and energy loss in conducting materials. This work is organized as follows: we first present the theory in Sect. II, details about our computational framework in Sect. III, the interpolation of electron, phonon and electron-phonon quantities in Sect. IV, the calculation of electron-phonon properties in Sect. V, and the final remarks and conclusions in Sect. VI.

Theory

We start with the definition of the EPME in the phonon mode representation [77]

gmnν(k,q)=∑καℏ2Mκωqνeκανq⟨ψmk+q|∂κα,qV|ψnk⟩(1)g_{mn\nu}(\mathbf{k},\mathbf{q}) = \sum_{\kappa\alpha} \sqrt{\frac{\hbar}{2M_{\kappa}\omega_{\mathbf{q}\nu}}}e_{\kappa\alpha\nu\mathbf{q}}\left\langle\psi_{m\mathbf{k}+\mathbf{q}}\middle|\partial_{\kappa\alpha,\mathbf{q}}V\middle|\psi_{n\mathbf{k}}\right\rangle \tag*{(1)}

where ωqν\omega_{\mathbf{q}\nu} is the frequency of the phonon mode ν\nu with quasi-momentum q\mathbf{q}, eκανqe_{\kappa\alpha\nu\mathbf{q}} the associated eigenvector, MκM_{\kappa} is the mass of the κ\kappa-th atom in the unit cell, and ∂κα,qV\partial_{\kappa\alpha,\mathbf{q}}V is the variation of the effective potential with respect to the perturbation involving atom κ\kappa displaced in Cartesian direction α\alpha. This definition has to be updated when considering non-collinear magnetism as

gmnνσσ′(k,q)=∑καℏ2Mκωqνeκανq⟨ψmk+qσ′|∂κα,qVσ′σ|ψnkσ⟩(2)g_{mn\nu}^{\sigma\sigma'}(\mathbf{k},\mathbf{q}) = \sum_{\kappa\alpha} \sqrt{\frac{\hbar}{2M_{\kappa}\omega_{\mathbf{q}\nu}}}e_{\kappa\alpha\nu\mathbf{q}}\left\langle\psi_{m\mathbf{k}+\mathbf{q}}^{\sigma'}\middle|\partial_{\kappa\alpha,\mathbf{q}}V^{\sigma'\sigma}\middle|\psi_{n\mathbf{k}}^{\sigma}\right\rangle \tag*{(2)}

where σ\sigma and σ′\sigma' are the spinor indices. The matrix element gmnνσσ′(k,q)g_{mn\nu}^{\sigma\sigma'}(\mathbf{k},\mathbf{q}) can be written in terms of the deformation potential Dmnνσσ′(k,q)D_{mn\nu}^{\sigma\sigma'}(\mathbf{k},\mathbf{q}) as [78]

∣gmnνσσ′(k,q)∣=ℏmuωqνDmnνσσ′(k,q),(3)\left|g_{mn\nu}^{\sigma\sigma'}(\mathbf{k},\mathbf{q})\right| = \sqrt{\frac{\hbar}{m_{\mathrm{u}}\omega_{\mathbf{q}\nu}}}D_{mn\nu}^{\sigma\sigma'}(\mathbf{k},\mathbf{q}), \tag*{(3)}

with mum_{\mathrm{u}} being the atomic mass constant. Dmnνσσ′(k,q)D_{mn\nu}^{\sigma\sigma'}(\mathbf{k},\mathbf{q}) gives a measure of how much a phonon perturbation has changed the internal self-consistent potential felt by the electrons. When taking the absolute value of ∣gmnνσ′σ(k,q)∣|g^{\sigma'\sigma}_{mn\nu}(\mathbf{k},\mathbf{q})|, we take the root mean square average over any degenerate manifolds in mm, nn and ν\nu. Note that the definitions of Dmnνσ′σ(k,q)D^{\sigma'\sigma}_{mn\nu}(\mathbf{k},\mathbf{q}) and gmnνσ′σ(k,q)g^{\sigma'\sigma}_{mn\nu}(\mathbf{k},\mathbf{q}) are also valid for the case of nonmagnetic compounds with and without SOC. The main property exploited in the nonmagnetic case are the following extra symmetries in the spin indices [gmnνσ′σ(k,q)]∗=(−1)σ+σ′gmnνσˉ′σˉ(−k,−q)[g^{\sigma'\sigma}_{mn\nu}(\mathbf{k},\mathbf{q})]^* = (-1)^{\sigma+\sigma'}g^{\bar{\sigma}'\bar{\sigma}}_{mn\nu}(-\mathbf{k},-\mathbf{q}), with σˉ\bar{\sigma} being the opposite spin to σ\sigma. This means that if σ=σ′\sigma= \sigma' then [gmnνσ′σ(k,q)]∗=gmnνσˉ′σˉ(−k,−q)[g^{\sigma'\sigma}_{mn\nu}(\mathbf{k},\mathbf{q})]^* = g^{\bar{\sigma}'\bar{\sigma}}_{mn\nu}(-\mathbf{k},-\mathbf{q}), and if σ≠σ′\sigma\ne\sigma' then [gmnνσ′σ(k,q)]∗=−gmnνσˉ′σˉ(−k,−q)[g^{\sigma'\sigma}_{mn\nu}(\mathbf{k},\mathbf{q})]^* = -g^{\bar{\sigma}'\bar{\sigma}}_{mn\nu}(-\mathbf{k},-\mathbf{q}) (e.g. for heavy elements with spin orbit coupling but no magnetization). These two results are a consequence of time-reversal symmetry, which is absent in magnetic compounds. In practice, performing magnetic calculations with SOC, commonly known as non-collinear magnetic calculations, is computationally expensive (typically by a factor of 23=82^3 = 8 for cubic scaling DFT) in systems with several atoms. The collinear approximation consists in separating the two spin channels into up (↑\uparrow) and down (↓\downarrow) without explicit interaction between them, thereby only doubling the cost of the calculations, regardless of the system size. This allows us to define a spin-resolved EPME:

gmnνσ(k,q)=δσ′σgmnνσ′σ(k,q)(4)g^\sigma_{mn\nu}(\mathbf{k},\mathbf{q}) = \delta_{\sigma'\sigma}g^{\sigma'\sigma}_{mn\nu}(\mathbf{k},\mathbf{q}) \tag*{(4)}

and it follows that the deformation potential becomes

Dmnνσ(k,q)=δσ′σDmnνσ′σ(k,q).(5)D^\sigma_{mn\nu}(\mathbf{k},\mathbf{q}) = \delta_{\sigma'\sigma}D^{\sigma'\sigma}_{mn\nu}(\mathbf{k},\mathbf{q}). \tag*{(5)}

In systems with many valence electrons, we are commonly interested in the deformation potential of several bands. For each spin channel, we can consider

Dνσ(k,q)=∑mn∈M[Dmnνσ(k,q)]2(6)D^\sigma_\nu(\mathbf{k},\mathbf{q}) = \sqrt{\sum_{mn\in\mathcal{M}}[D^\sigma_{mn\nu}(\mathbf{k},\mathbf{q})]^2} \tag*{(6)}

as the deformation potential of interest, with the sum over the electronic bands m,nm,n restricted to the bands in the Wannier manifold M\mathcal{M}. We also extend the mode- and momentum-resolved electron-phonon coupling λqν\lambda_{\mathbf{q}\nu}, related to the imaginary part of the phonon self energy Im⁡Πqν\operatorname{Im}\Pi_{\mathbf{q}\nu}, which in the double delta approximation at the collinear level becomes [79]

λqν≡∑σmnkλqνσ=∑σmnkwkN(εF)ωqν∣gmnνσ(k,q)∣2×δ(εF−εnkσ)δ(εF−εmk+qσ)=1πN(εF)Im⁡Πqνωqν2,(7)\begin{aligned} \lambda_{\mathbf{q}\nu} \equiv\sum_{\sigma mn\mathbf{k}}\lambda^\sigma_{\mathbf{q}\nu} = \sum_{\sigma mn\mathbf{k}}\frac{w_{\mathbf{k}}}{N(\varepsilon_{\mathrm{F}})\omega_{\mathbf{q}\nu}}|g^\sigma_{mn\nu}(\mathbf{k},\mathbf{q})|^2 \\ &\times\delta(\varepsilon_{\mathrm{F}}-\varepsilon^\sigma_{n\mathbf{k}})\delta(\varepsilon_{\mathrm{F}}-\varepsilon^\sigma_{m\mathbf{k}+\mathbf{q}}) = \frac{1}{\pi N(\varepsilon_{\mathrm{F}})}\frac{\operatorname{Im}\Pi_{\mathbf{q}\nu}}{\omega_{\mathbf{q}\nu}^2}, \tag*{(7)} \end{aligned}

where N(εF)N(\varepsilon_{\mathrm{F}}) is the total density of states at the Fermi energy εF\varepsilon_{\mathrm{F}} and wkw_{\mathbf{k}} are the k-point integration weights normalized to one. In the non-magnetic case, the sum over σ\sigma is removed and ∑kwk=2\sum_{\mathbf{k}}w_{\mathbf{k}}=2 due to time-reversal symmetry. We note that another definition exists for the spin-resolved electron-phonon coupling using the partial density of states at the Fermi level Nσ(εF)N^\sigma(\varepsilon_{\mathrm{F}}) [60, 80]. However, this convention will suffer from divergences in half-

and semi-metals, where only one spin channel contributes to the total density of states. The present version should be used systematically for spin polarized cases, taking into account the factor of 2 between NσN^\sigma and NN in the unpolarized case.

Another important quantity is the mode- and momentum-resolved phonon linewidth γqν\gamma_{\mathbf{q}\nu}

γqν≡∑σγqνσ=2πωqν∑σkmnwk∣gmnνσ(k,q)∣2×δ(εF−εnkσ)δ(εF−εmk+qσ),(8)\begin{aligned} \gamma_{\mathbf{q}\nu} \equiv\sum_\sigma\gamma^\sigma_{\mathbf{q}\nu}=2\pi\omega_{\mathbf{q}\nu}\sum_{\sigma kmn}w_{\mathbf{k}}|g^\sigma_{mn\nu}(\mathbf{k},\mathbf{q})|^2 \\ &\times\delta(\varepsilon_{\mathrm{F}}-\varepsilon^\sigma_{n\mathbf{k}})\delta(\varepsilon_{\mathrm{F}}-\varepsilon^\sigma_{m\mathbf{k}+\mathbf{q}}), \tag*{(8)} \end{aligned}

which can be used to compute the Eliashberg spectral function

α2F(ω)=12πN(εF)∑σqνγqνσωqνδ(ω−ωqν),(9)\alpha^2F(\omega)=\frac{1}{2\pi N(\varepsilon_{\mathrm{F}})}\sum_{\sigma\mathbf{q}\nu}\frac{\gamma^\sigma_{\mathbf{q}\nu}}{\omega_{\mathbf{q}\nu}}\delta(\omega-\omega_{\mathbf{q}\nu}), \tag*{(9)}

where wqw_{\mathbf{q}} are the q\mathbf{q}-point integration weights normalized to one. The Eliashberg spectral function can be used to determine the contribution of specific phonon modes to the total electron-phonon coupling strength defined as

λ≡∫dωωα2F(ω),(10)\lambda\equiv\int\frac{d\omega}{\omega}\alpha^2F(\omega), \tag*{(10)}

This quantity provides a measure of the overall mass enhancement in a metallic system.

Finally, we study resistive transport in magnetic metals. The transport of charge in a magnetic material depends on the spin-resolved carrier scattering rate 1/τnkσ1/\tau^\sigma_{n\mathbf{k}} defined as

1τnkσ≡∑qmνwq∣gmnνσ(k,q)∣2×[(nqν+1−fmk+qσ)δ(εnkσ−εmk+qσ−ℏωqν)\begin{aligned} \frac{1}{\tau^\sigma_{n\mathbf{k}}}\equiv\sum_{\mathbf{q}m\nu}w_{\mathbf{q}}|g^\sigma_{mn\nu}(\mathbf{k},\mathbf{q})|^2 \\ &\times\left[(n_{\mathbf{q}\nu}+1-f^\sigma_{m\mathbf{k}+\mathbf{q}})\delta(\varepsilon^\sigma_{n\mathbf{k}}-\varepsilon^\sigma_{m\mathbf{k}+\mathbf{q}}-\hbar\omega_{\mathbf{q}\nu})\right. \end{aligned}
+(nqν+fmk+qσ)δ(εnkσ−εmk+qσ+ℏωqν)].(11)\left.+(n_{\mathbf{q}\nu}+f^\sigma_{m\mathbf{k}+\mathbf{q}})\delta(\varepsilon^\sigma_{n\mathbf{k}}-\varepsilon^\sigma_{m\mathbf{k}+\mathbf{q}}+\hbar\omega_{\mathbf{q}\nu})\right]. \tag*{(11)}

where wqw_{\mathbf{q}} follow the same integration rules as wkw_{\mathbf{k}}, fmk+qσf^\sigma_{m\mathbf{k}+\mathbf{q}} is the Fermi-Dirac occupation for spin σ\sigma and nqνn_{\mathbf{q}\nu} is the Bose-Einstein distribution function. As for previous quantities, the average scattering rate is additive in the collinear approximation to magnetism:

⟨τ−1⟩∼∑σ∑nkwk1τnkσ.(12)\langle\tau^{-1}\rangle\sim\sum_\sigma\sum_{n\mathbf{k}}w_{\mathbf{k}}\frac{1}{\tau^\sigma_{n\mathbf{k}}}. \tag*{(12)}

Previous studies have used the lowest-order variational approximation (LOVA) to study the resistivity of magnetic systems [49], which is based on computing the average scattering rate. This is a strong approximation in anisotropic systems or systems with different Fermi sheets [81]. Here, we solve the linearized Boltzmann transport equation (BTE) to calculate the conductivity [82, 83]

σαβσ=−eVuc∑nkwkvnkασ∂εβfnkσ,(13)\sigma^\sigma_{\alpha\beta}=-\frac{e}{V_{\mathrm{uc}}}\sum_{n\mathbf{k}}w_{\mathbf{k}}v^\sigma_{n\mathbf{k}\alpha}\partial_{\varepsilon_\beta}f^\sigma_{n\mathbf{k}}, \tag*{(13)}

where VucV^{\mathrm{uc}} is the volume of the unit cell, ∂Eβfnkσ\partial_{E_\beta}f^\sigma_{n\mathbf{k}} is the perturbed occupation factor due to an applied electric field EβE_\beta in the direction β\beta, (see, Eq. (16) below) vnkασv^\sigma_{n\mathbf{k}\alpha} is the α\alpha component of the spin-resolved electronic band velocity, and σαβσ\sigma^\sigma_{\alpha\beta} is the conductivity tensor for spin index σ\sigma. The corresponding resistivity is

ραβBTE=(σαβ↑+σαβ↓)−1=[(ραβ↑)−1+(ραβ↓)−1]−1.(14)\rho^{\mathrm{BTE}}_{\alpha\beta} = \left(\sigma^\uparrow_{\alpha\beta}+\sigma^\downarrow_{\alpha\beta}\right)^{-1} = \left[\left(\rho^\uparrow_{\alpha\beta}\right)^{-1}+\left(\rho^\downarrow_{\alpha\beta}\right)^{-1}\right]^{-1}. \tag*{(14)}

which is the usual law of addition for resistivities in parallel, and in this case ραβσ\rho^\sigma_{\alpha\beta} is the resistivity experienced by the electrons in each spin channel. Interestingly, when using the Ziman or LOVA formula, the resistivity becomes additive instead ραβ↑+ραβ↓\rho^\uparrow_{\alpha\beta}+\rho^\downarrow_{\alpha\beta}, questioning the validity of Eq. (8) of Ref. [49], see Supplementary Material S1 for more information. When comparing to polycrystalline samples, we use

ρBTE=Tr⁡[ραβBTE]3.(15)\rho^{\mathrm{BTE}} = \frac{\operatorname{Tr}\left[\rho^{\mathrm{BTE}}_{\alpha\beta}\right]}{3}. \tag*{(15)}

The out-of-equilibrium occupations ∂Eβfnkσ\partial_{E_\beta}f^\sigma_{n\mathbf{k}} [81] in the collinear magnetic case are obtained by solving the BTE

∂Eβfnkσ=evnkβσ∂fnkσ∂εnkστnkσ+2πτnkσℏ∑mνqwq∣gmnν(k,q)∣2×[(nqν+1−fnkσ)δ(εnkσ−εmk+qσ−ℏωqν)+(nqν+fnkσ)δ(εnkσ−εmk+qσ+ℏωqν)]∂Eβfmk+qσ.(16)\begin{aligned} \partial_{E_\beta}f^\sigma_{n\mathbf{k}} = e v^\sigma_{n\mathbf{k}\beta}\frac{\partial f^\sigma_{n\mathbf{k}}}{\partial\varepsilon^\sigma_{n\mathbf{k}}}\tau^\sigma_{n\mathbf{k}} + \frac{2\pi\tau^\sigma_{n\mathbf{k}}}{\hbar}\sum_{m\nu\mathbf{q}}w_{\mathbf{q}}\left|g_{mn\nu}(\mathbf{k},\mathbf{q})\right|^2 \\ &\times\left[(n_{\mathbf{q}\nu}+1-f^\sigma_{n\mathbf{k}})\delta(\varepsilon^\sigma_{n\mathbf{k}}-\varepsilon^\sigma_{m\mathbf{k}+\mathbf{q}}-\hbar\omega_{\mathbf{q}\nu}) +(n_{\mathbf{q}\nu}+f^\sigma_{n\mathbf{k}})\delta(\varepsilon^\sigma_{n\mathbf{k}}-\varepsilon^\sigma_{m\mathbf{k}+\mathbf{q}}+\hbar\omega_{\mathbf{q}\nu})\right]\partial_{E_\beta}f^\sigma_{m\mathbf{k}+\mathbf{q}}. \tag*{(16)} \end{aligned}

Here ∂fnkσ/∂εnkσ\partial f^\sigma_{n\mathbf{k}}/\partial\varepsilon^\sigma_{n\mathbf{k}} is the derivative of the Fermi-Dirac distribution. The equation is solved through an iterative procedure and if we neglect the second term in Eq. (16), we obtain the self-energy relaxation time approximation (SERTA) [82], which in the case of metallic systems is usually quite accurate.

Methods

DFT calculations are performed using QE [62, 63] version 7.4 and Abinit version 10.2.5 [58, 59, 61, 84]. Scalar-relativistic norm-conserving pseudopotentials [85] from the standard accuracy table of the PseudoDojo project [86] are used. We use the local density approximation (LDA) [87] and the generalized gradient approximation (GGA) with the Perdew Burke Ernzerhof (PBE) functional [88]. The interpolated phonon dispersions and dynamical matrices are obtained with DFPT [42–45]. The maximally localized Wannier functions are obtained using the wannier90 package version 3.1 [89]. Direct evaluation of spin-resolved EPME is obtained by modifying QE to print the values. We provide the modified code on the Materials Cloud Archive [90] along with all data to reproduce the results presented here. The interpolation of electronic band structures, phonon band structures, and electron phonon matrix elements, as well

as the calculation of transport properties and electron-phonon coupling matrix elements are performed by modifying the EPW package version 5.9 [70]. This feature will be released in a future EPW version and we also provide the code on the Materials Cloud Archive [90]. The cutoff energy for the plane wave expansion, k\mathbf{k}-meshes for the electrons, and the q\mathbf{q}-meshes for the phonons are selected so that an arbitrary increase in those parameters produces a total energy change of less than 1 meV per unit cell. This corresponds to 100 Ry for both Fe and Ni, as well as a k\mathbf{k}-point grid of 24324^3 and a q\mathbf{q}-point grid of 838^3 for both Ni and Fe. Structural relaxations are performed with a convergence criterion of ΔE<10−12\Delta E < 10^{-12} Ry, and with a pressure of less than 1 bar. In the case of phonons, the convergence criteria in order to obtain accurate force constants was tr2_ph = 10−2010^{-20} for QE while for Abinit the convergence criteria is on the energy ΔE<10−20\Delta E < 10^{-20} Ha. We converge the resistivity below 1% above 150 K using EPW, and validate the results with an in-house developed version of the Electron-Phonon.jl package [91, 92], which relies on the real-space EPME computed with the EPW code [69, 70] that we extend to the collinear magnetic case for the purpose of this work. Excellent agreement is obtained between the two codes, see Supplementary Material S2 [90] for more information.

Electron-phonon coupling with magnetism

In this work, we study BCC Fe and FCC Ni, both in their ferromagnetic configuration using the collinear approximation. We also consider non-magnetic Pb and verify that the LSDA implementation yields the same results as the LDA case as described in Supplementary Material S3 [90].

Verification of the direct evaluation of the electron-phonon coupling

To verify our implementation, we compare the QE results with those obtained with ABINIT. Although both codes match well in non-magnetic compounds [71–73], no validation study had yet been performed in the magnetic case. To this end, we compare basic quantities including total energy, total magnetization and phonon frequencies while keeping the computational parameters as close as possible. We find an excellent level of agreement between the two codes that is similar to the non-magnetic case, see Supplementary Material S4 [90].

We compare the absolute value of the deformation potentials between the two codes. To achieve this, we note that QE performs an average over all degenerate states in electronic and phonon band indices. We therefore applied the same average to the ABINIT results and present a comparison for ferromagnetic Fe in Figure 1 between the two codes for k=Γ\mathbf{k}=\Gamma and bands m=n=5m=n=5. We find a good agreement between the two codes with a maximum error of less than 0.02 meV/bohr. We provide the same comparison for Ni in Supplemental Material S5 [90] where we find an error below 0.01 meV/bohr.

Direct evaluation of ferromagnetic iron deformation potential for spin up and spin down

Figure 1. Direct evaluation of ferromagnetic iron deformation potential Dm=5,n=5ν(k=Γ,q)D^{\nu}_{m=5,n=5}(\mathbf{k}=\Gamma,\mathbf{q}) for (a) spin up and (b) spin down at k=Γ\mathbf{k}=\Gamma for m=n=5m=n=5 along a q\mathbf{q} momentum line from Γ\Gamma to N for all phonon modes ν\nu (LA and TA). Red crosses are computed with QE while blue squares are computed with Abinit. The LDA functional is used.

Verification of the electron phonon interpolation

Since we obtained reliable EPME by direct evaluation in QE, we use the EPW package to interpolate them into finer q\mathbf{q} and k\mathbf{k} grids. We extend the EPW package to be compatible with collinear magnetism and introduce two new input variables to control the selection of spin channel.

We first compare in Figure 2(a) and 2(b) the direct and Wannier-interpolated electronic band structures of ferromagnetic Fe and Ni respectively, obtaining an energy mismatch of less than 0.1 meV for all the displayed k\mathbf{k}-points and bands. This shows that both spin channels are correctly reproduced by the interpolation procedure. We then compute in Figure 2(c) and 2(d) the phonon dispersion of ferromagnetic Fe and Ni respectively using LDA and PBE. Overall, LDA overestimates by about 10% the phonon frequency with respect to neutron scattering experiments [93, 94]. Using the PBE functional yields a much better agreement, in line with a previous study [95]. Moreover, we validate the interpolated electron-phonon deformation potential by comparing to direct calculations in Figure 3, for electronic bands crossing the Fermi level and at the electronic zone center. We show the two spin channels for Fe and Ni, and find that convergence is achieved for 12×12×1212 \times12 \times12 k\mathbf{k}/ q\mathbf{q} coarse grids in both cases. For such grids, we find excellent agreement between interpolated and independently calculated reference direct calculations. In addition, we also compare the EPW interpolation with the Abinit interpolation finding similar results; see Supplementary Material S5 [90]. Finally, we verify that the Hamiltonian, dynamical matrices, and EPME all decay rapidly in real space; see Supplementary

Electronic band structures and phonon band structures for ferromagnetic Fe and Ni

Figure 2. Electronic band structure for ferromagnetic (a) Fe and (b) Ni. Solid (dashed) lines correspond to QE (EPW) calculations. Red/orange (blue/cyan) lines correspond to spin ↑ (↓) or majority (minority) channel. Phonon band structure for (c) Fe and (d) Ni with LDA (green) and PBE (orange) functionals where the black dots are experimental values from Refs. [93, 94].

Spin resolved deformation potential for Fe and Ni

Figure 3. Spin resolved deformation potential for the three phonon branches ν\nu (LA and TA) at k=Γ\mathbf{k}=\Gamma and as a function of wave vector q\mathbf{q}. The color scheme is presented as an inset in the figure ↑ (red/orange/yellow/green) and ↓ (blue/pink/purple). The solid dots are direct evaluation of the deformation potential with QE. The lines are the EPW interpolated deformation potential from 434^3, 636^3, 838^3, 10310^3 and 12312^3 k/q\mathbf{k}/\mathbf{q} homogeneous coarse grids as presented in the inset of the figure. The electronic bands mm and nn are summed using Eq. (6) from second to sixth lowest electronic bands shown in Figure 2(a) and 2(b) for (a) Fe and (b) Ni.

Material S6 [90].

Physical properties and electron-phonon quantities

Having validated our spin-resolved Wannier-interpolation scheme in magnetic materials, we compute physical observables linked with electron-phonon interactions. We start by computing the full electron-phonon coupling strength λ\lambda, and the corresponding spin-resolved λσ\lambda^{\sigma} using Eq. (7)

λσ=∑qνwqλqνσ(17)\lambda^{\sigma} = \sum_{q\nu} w_q \lambda_{q\nu}^{\sigma} \tag*{(17)}

and we report the results for ferromagnetic Fe and Ni in Table I. Interestingly, we find that the calculated spin-resolved electron phonon coupling strength shows large differences between the two spin channels in both Fe and Ni. Although this behavior is not entirely surprising for Fe given the clear difference in deformation potential between the two spin channels, the same cannot be said in the case of Ni, whose splitting in deformation potential is much weaker, likely due to its smaller exchange splitting. Additionally, we note that the main contribution to the electron phonon coupling strength in the case of Fe comes from the spin up (majority) channel, while in the case of Ni we find the reverse situation. In order to understand this behavior, we inspect the definition of the spin-resolved electron-phonon coupling strength Eqs. (7) and (17) and we notice two main sources for this discrepancy. The first one is the deformation potential. The second is the electronic nesting for the given spin channel provided by the product of the two delta functions. This quantity is related to the density of states at the Fermi level. As seen in Figure 3, the deformation potential is very different between the two spin channels for both Fe and Ni. In the case of Fe the difference in the deformation potential is significant and could explain the difference. However, λ↓\lambda^{\downarrow} is smaller than λ↑\lambda^{\uparrow}, while the trend in the deformation potential is opposite, see Figure 3(a). Consequently, the deformation potential cannot be the source of the difference between λ↑\lambda^{\uparrow} and λ↓\lambda^{\downarrow}. In the case of Ni, both spin channels have similar values in the deformation potential, while

CompoundMagnetismλ\lambdaλ↑\lambda^{\uparrow}λ↓\lambda^{\downarrow}
Fe (BCC)FM0.29610.21800.0781
Ni (FCC)FM0.27670.00400.2727

Table I. Total and spin-resolved electron phonon coupling strength λ\lambda for ferromagnetic (FM) Fe and Ni.

the electron phonon coupling strength λ\lambda is completely dominated by the spin down channel λ≈λ↓\lambda\approx\lambda^{\downarrow}. Therefore, the deformation potential is not the main driver for such differences in either of the two considered cases.

Inspecting the density of states close to the Fermi level for each compound in Figure 4, we note that for Fe the spin up channel (majority) is 2.3 times larger than the spin down channel (minority), see Figure 4(a), while the opposite behavior is found for Ni where the spin down (minority) channel is 7.6 times larger, see Figure 4(b). This is a well-known effect linked to the dd-band filling. Thus, the differences in the spin-resolved electron-phonon coupling strength λσ\lambda^{\sigma} between the two spin channels for Fe and Ni are mainly related to Fermi-level nesting and the density of states at the Fermi level. Our results for the electron-phonon coupling strength of Fe compare well with those obtained in [79].

Spin resolved density of states for ferromagnetic Fe and Ni

Figure 4. Spin resolved density of states for ferromagnetic (a) Fe and (b) Ni. Positive and negative density of states representing up (red) and down (blue) spin channels.

We also compute the Eliashberg spectral functions and show the result in Figure 5. We see in Figure 5(a) that the Fe spin majority channel is the dominant contribution for most of the energy range. In contrast, for Ni the spin minority dominates throughout and the spin majority is negligible, see Figure 5(b). This important difference is due to the density of states at the Fermi level in Ni, see Figure 4. Recently, some authors have explored phonon-mediated superconductivity in magnetic systems by means of the mass enhancement λ\lambda and the Eliashberg spectral function [55]. Although the validity of using the Allen-Dyne formula [96] in magnetic materials has not yet been demonstrated, we report superconducting results for the non-magnetic and magnetic cases in Supplementary Material S7 [90]. In both cases and for both materials, we find a vanishing superconducting temperature in agreement with experimental observation.

Eliashberg spectral function for ferromagnetic Fe and Ni

Figure 5. Eliashberg spectral function for ferromagnetic (a) Fe and (b) Ni. Red lines corresponds to the spin up (majority) and blue line corresponds to spin down (minority).

Resistivity of magnetic metals

We now focus on the resistivity of metals and solve the BTE for both spin channels in ferromagnetic Fe and Ni. We note that when converging the resistivity and in general electron-phonon related quantities, it is possible to use different k/q\mathbf{k}/\mathbf{q} coarse and fine grids for the two spin channels since there is no guarantee that both spin channels would behave equivalently. Nevertheless, in our case, we find that convergence is achieved with the same settings. A good compromise between accuracy and computational cost is obtained with a 838^{3} k/q coarse grid; see Supplementary Material S8 [90] for further discussion.

The computed resistivity of Fe is presented in Figure 6(a) and shows that the BTE with electron-phonon scattering explains only partially the resistivity and tends to underestimate it. We find that PBE results are slightly closer to experiment due to the improved phonon dispersion, see Figure 2(c), and a small decrease in Fermi velocity; see Supplementary material S9 [90]. We also find that the SERTA is a good approximation to the iterative BTE solutions, which is typical for non-magnetic metals [97]. Our results are qualitatively similar to a previous study using the Ziman formula [49] where quantitative agreement with the experimental results is obtained at low temperatures. However, we notice that the electron-phonon contribution to resistivity does seem to dominate at low temperature, see inset in Figure 6(a), and at temperatures up to 300 K our calculations suggest that electron-phonon scattering can account for at least 75% of the experimental resistivity when using the PBE functional. For higher temperatures, the observed underestimation indicates that electron-phonon scattering is not the sole mechanism contributing to the resistivity.

Figure 6. Electrical resistivity as a function of temperature for (a) Fe and (b) Ni, with an inset at low temperature 0-200 K. Black symbols are obtained from Refs. [98–100]. Lines are computed resistivities using SERTA (dashed) and full BTE (solid) with LDA (green) and PBE (orange).

We continue our analysis by investigating the case of ferromagnetic Ni. As seen in Figure 6(b), the computed resistivity for Ni is also underestimated in the entire temperature range compared to experiment. However, above 350 K the resistivity originating from electron-phonon scattering is almost one order of magnitude less than the experimental resistivity. In addition, the agreement at low temperatures is worse than in Fe since there is an important underestimation of the resistivity already in the 50-100 K temperature range, see the inset in Figure 6(b). Interestingly, we note that using the PBE functional does not improve significantly in spite of the remarkably better agreement of the phonon dispersion with experiments, see Figure 2(d). This hints at an important distinction with respect to Fe.

In ferromagnetic compounds, contributions to the resistivity originating from magnons or spin fluctuations are expected to be significant [101, 102]. One of the reasons why magnons may explain the missing resistivity of Ni and Fe above 300 K is due to their energy scale which is around 8 to 9 times bigger than the phonons [103]. The magnon population scales as (T/TCurie)2(T/T_{\mathrm{Curie}})^2 [102], and will be smaller in Fe than in Ni for a given temperature due to the lower Curie temperature (TCurieT_{\mathrm{Curie}}) of 627 K [99] in Ni, compared to 1043 K in Fe [100]. Consequently, while at 300 K the phonon contribution to the resistivity would show a linear “high temperature behavior”, the magnon contribution would still be within its “low temperature regime” with a T2T^2 behavior. Although this explains the increase in the importance of magnons at temperatures beyond 200-300 K, it does not explain the marked difference between Fe and Ni.

In fact, the magnon population is crucially dependent on the density of states at the Fermi level of the opposite spin channel [104]. The large density of states of the spin down (minority) channel in Ni (see Figure 4(b)) above the Fermi level allows for a large magnon creation by converting spin up (majority) electrons into spin down (minority) electrons. In the case of Fe, magnon creation is suppressed by the small accessible phase space of minority electrons [104]. In addition to the difference in magnon creation, our calculations show that the spin up electrons in Ni experience far less scattering from electron-phonon, and, as a result, our calculations show that ραβ≈ραβ↑\rho_{\alpha\beta}\approx\rho_{\alpha\beta}^{\uparrow} with an order of magnitude difference in the conductivities σαβ↑≫σαβ↓\sigma_{\alpha\beta}^{\uparrow}\gg\sigma_{\alpha\beta}^{\downarrow}, while for Fe ραβ≈ραβ↓\rho_{\alpha\beta}\approx\rho_{\alpha\beta}^{\downarrow}. Since the spin up channel for Ni shows a far larger lifetime broadening due to magnons [104], the total lifetime τnk↑\tau_{nk}^{\uparrow} is smaller than with electron-phonon scattering only, explaining the underestimated resistivity compared to experiment in Ni. In summary, comparing the calculated resistivity at low temperature with the experimental data of Fe and Ni, we find that the behavior of these two compounds is fundamentally different: magnons dominate the scattering in Ni while electron-phonon is the dominant scattering mechanism in Fe up to room temperature. We also consider the case of ferromagnetic Co HCP finding that the resistivity would be a case in between Fe and Ni; see Supplementary material S10 [90]

Role of magnetism on physical properties

One of the key aspects of including magnetism is the changes on both electronic and phonon band dispersion. Therefore, we consider the effects of using the non-magnetic (NM) approximation in Fe and Ni by relaxing their BCC and FCC structures and calculating their respective phonon dispersions, see Figure 7. As we can see in Figure 7(a), Fe is not dynamically stable when we use the NM approximation. This is due to a fermionic pressure that increases the lattice parameters of Fe when magnetism is allowed [105–108]. These instabilities found in the phonon dispersion for a BCC structure in the NM approximation are known to be related to distortions that lead to hexagonal phases [109]. This is also consistent with the fact that in the crystallographic phase diagram of Fe, we find a hexagonal phase at pressures a bit above 10 GPa [110, 111]. Interestingly, the phonon dispersion for Ni does not change much by using the NM approximation; see Figure 7(b). In fact, one can notice that the FM phase of Ni produces only a small overall imbalance in the spin occupations since the experimental magnetic moment per unit cell is Mexp=0.616μBM_{\mathrm{exp}} = 0.616 \mu_{\mathrm{B}} [112] when compared with Fe that presents a magnetic moment of Mexp=(2.2±0.1)μBM_{\mathrm{exp}} = (2.2 \pm0.1) \mu_{\mathrm{B}} [113–115]. Thus, one may wonder how different the resistivity in the NM phase is, when compared with that in the FM collinear case for Ni. In that regard, we repeat the calculations with the BTE for the resistivity of Ni with PBE (see LDA in the supplementary material S11 [90]) but using the NM phase and we compare it in Figure 8(b) with the FM case and with the experimental data. We find that the NM approximation produces a slight overestimation of the resistivity for temperatures below 300 K and then it underestimates the resistivity.

Phonon dispersion for spin-unpolarized Fe and Ni

Figure 7. Phonon dispersion for spin-unpolarized (non-magnetic) (a) Fe and (b) Ni obtained with PBE functional. Black dots are the experimental data from Refs. [93, 94].

Electrical resistivity of iron and nickel as a function of temperature

Figure 8. Electrical resistivity of iron and nickel as a function of temperature. The solid green and purple lines are the computed PBE resistivities ρBTE\rho^{\mathrm{BTE}} using ferromagnetic and non-magnetic EPC, respectively. Black dots are experimental values obtained from Refs. [98–100].

Although the limitation of the NM approximation has been highlighted previously [116–120], it is common to

find studies that still use the NM approximation to model spin-disordered paramagnetic, the ferromagnetic or antiferromagnetic phases [121–123]. In addition, some studies have also neglected the soft modes and still perform transport calculations [75, 76, 124] For this reason, and in order to highlight how bad it would be to perform these calculations, we also compute the resistivity of NM Fe as a function of temperature and compare it with the FM case in Figure 8(a). The NM calculation with negative phonon frequencies yield an extremely overestimated resistivity when compared to the FM one and the experimental results. Importantly, we show that even though the phonons are similar in Ni, neglecting the spin degree of freedom yields the wrong conclusion that the resistivity is dominated by electron-phonon interactions. We remark that, in spite of the seemingly better agreement with the experiment, the NM results actually overestimate the experimental data, which should never happen since we are not considering the full set of scattering channels. Moreover, the temperature-dependent trend of the NM result is inconsistent with experimental observation.

Conclusions

We explore the role of electron-phonon interactions in typical magnetic metals. Namely, ferromagnetic Fe and Ni, where we show that neither the deformation potential nor the electron-phonon coupling constant present the same strength for the two spin channels, with the latter being dominated by Fermi surface nesting effects and the partial density of states of each spin channel. Additionally, we show that the electrical resistivity of Fe and Ni exhibits important contributions from electron-phonon scattering at room temperature although the relative importance is material dependent. While in Fe the resistivity due to electron-phonon interactions is the dominant contribution at room temperature, for Ni it is less than one-third of the total. Additionally, we show that including the spin degree of freedom to compute resistivity is crucial. Neglecting it can lead to unstable phonons such as in the case of Fe, can overestimate resistivity, and can yield a completely erroneous interpretation for the source of the resistivity. Furthermore, we explore the phonon-mediated superconducting mechanism in these two elemental compounds, a subject rarely addressed with proper numerical calculations. We find suppressed superconductivity, showing that even in a non-magnetic configuration their critical temperatures would be extremely low. These results validate and extend the capabilities of EPW to compute electron-phonon quantities with collinear magnetism. Beyond the methodological advances, our findings highlight how magnetism critically shapes resistive losses and superconductivity in metals that are central to energy applications. By enabling accurate modeling of transport in magnetic systems, this work provides a foundation for the design of low-energy-consumption spintronic devices, improved conductive pathways with reduced losses, and future magnetic materials for energy conversion and transmission. Future extensions to include spin–orbit coupling and non-collinear magnetism will further expand the reach of this framework, opening new directions for first-principles studies of complex magnetic materials.

Acknowledgments

S. P. is a Research Associate and M.M. is an Aspirant PhD candidate of the Fonds de la Recherche Scientifique - FNRS. This work was supported by the Fonds de la Recherche Scientifique - FNRS under Grants number T.0183.23 (PDR). M.J.V. and G.A. acknowledge the Fonds de la Recherche Scientifique (FRS-FNRS Belgium) and Fonds Wetenschappelijk Onderzoek (FWO Belgium) for EOS project CONNECT (G.A. 40007563), and Fédération Wallonie Bruxelles and ULiege for funding ARC project DREAMS (G.A. 21/25-11). M.J.V. acknowledges funding by the Dutch Gravitation program “Materials for the Quantum Age” (QuMat, reg number 024.005.006), financed by the Dutch Ministry of Education, Culture and Science (OCW). Computational resources have been provided by the supercomputing facilities of the Université catholique de Louvain (CISM/UCL) and the Consortium des Equipements de Calcul Intensif en Fédération Wallonie Bruxelles (CÉCI) funded by the Fond de la Recherche Scientifique de Belgique (F.R.S.-FNRS) under convention 2.5020.11 and by the Walloon Region. The present research benefited from additional computational resources made available on Lucia, the Tier-1 supercomputer of the Walloon Region, infrastructure funded by the Walloon Region under the grant agreement n°1910247, and by EuroHPC (Extreme grant EHPC-EXT-2023E02-050) on Marenostrum5 at BSC, Spain.

References

Supplementary Material: Electron-phonon coupling in magnetic materials using the local spin density approximation

Álvaro Adrián Carrasco Álvarez; Matteo Giantomassi; Jae-Mo Lihm; Guillaume E. Allemand; Maxime Mignolet; Matthieu Verstraete; Samuel Poncé

LOVA RESISTIVITY FORMULA

One of the first formulas for computing the resistivity of metals due to electron phonon is the one derived by Ziman [1] and later reformulated by Grimvall [2] that reads

ρLOVA=4πmenee2kBT∫dωℏωα2Ftr(ω)n(ω,T)[n(ω,T)+1](18)\rho^{\mathrm{LOVA}} = \frac{4\pi m_e}{n_e e^2 k_{\mathrm{B}} T} \int d\omega\hbar\omega\alpha^2 F_{\mathrm{tr}}(\omega)n(\omega,T)[n(\omega,T)+1] \tag*{(18)}

with mem_e the mass of the electrons, n(ω,T)n(\omega,T) the Bose-Einstein occupation factor, and α2Ftr(ω)\alpha^2 F_{\mathrm{tr}}(\omega) being the so called transport spectral function defined as:

α2Ftr(ω)=1N(εF)∑kq∑mnνwkwq∣gmnν(k,q)∣2δ(ω−ωqν)[1−vnk⋅vmk+q∣vnk∣]δ(εnk−εF)δ(εmk+q−εF),(19)\alpha^2 F_{\mathrm{tr}}(\omega) = \frac{1}{N(\varepsilon_{\mathrm{F}})} \sum_{\mathbf{kq}} \sum_{mn\nu} w_{\mathbf{k}}w_{\mathbf{q}} \lvert g_{mn\nu}(\mathbf{k},\mathbf{q})\rvert^2 \delta(\omega-\omega_{\mathbf{q}\nu}) \left[1-\frac{\mathbf{v}_{n\mathbf{k}}\cdot\mathbf{v}_{m\mathbf{k+q}}}{\lvert\mathbf{v}_{n\mathbf{k}}\rvert}\right]\delta(\varepsilon_{n\mathbf{k}}-\varepsilon_{\mathrm{F}})\delta(\varepsilon_{m\mathbf{k+q}}-\varepsilon_{\mathrm{F}}), \tag*{(19)}

where N(εF)=∑n∫d3kΩBZδ(εnk−εF)N(\varepsilon_{\mathrm{F}})=\sum_n\int\frac{d^3\mathbf{k}}{\Omega_{\mathrm{BZ}}}\delta(\varepsilon_{n\mathbf{k}}-\varepsilon_{\mathrm{F}}) is the total density of states at the fermi level. Interestingly, this formula accounts for spin orbit coupling (SOC) in the Hamiltonian with ∑kwk=∑qwq=1\sum_{\mathbf{k}}w_{\mathbf{k}}=\sum_{\mathbf{q}}w_{\mathbf{q}}=1. If we neglect SOC, the Hamiltonian is diagonal and the spin index σ\sigma is a good quantum number and we obtain:

α2Ftr(ω)=∑σ1N(εF)∑kq∑mnνwkwq∣gmnνσ(k,q)∣2(1−vnkσ⋅vmk+qσ∣vnkσ∣)δ(εnkσ−εF)δ(εmk+qσ−εF)δ(ω−ωqν),(20)\alpha^2 F_{\mathrm{tr}}(\omega) = \sum_{\sigma}\frac{1}{N(\varepsilon_{\mathrm{F}})}\sum_{\mathbf{kq}}\sum_{mn\nu}w_{\mathbf{k}}w_{\mathbf{q}}\lvert g^\sigma_{mn\nu}(\mathbf{k},\mathbf{q})\rvert^2 \left(1-\frac{\mathbf{v}^\sigma_{n\mathbf{k}}\cdot\mathbf{v}^\sigma_{m\mathbf{k+q}}}{\lvert\mathbf{v}^\sigma_{n\mathbf{k}}\rvert}\right)\delta(\varepsilon^\sigma_{n\mathbf{k}}-\varepsilon_{\mathrm{F}})\delta(\varepsilon^\sigma_{m\mathbf{k+q}}-\varepsilon_{\mathrm{F}})\delta(\omega-\omega_{\mathbf{q}\nu}), \tag*{(20)}

where the nn and mm band indices are restricted to the specific spin subspace and ∑n→∑nσ\sum_n\rightarrow\sum_{n\sigma}. We can then define a spin resolved transport spectral function α2Ftrσ\alpha^2 F^\sigma_{\mathrm{tr}} such that α2Ftr(ω)\alpha^2 F_{\mathrm{tr}}(\omega) is additive in the spin indices σ\sigma

α2Ftr(ω)=∑σα2Ftrσ(ω)=α2Ftr↑(ω)+α2Ftr↓(ω).(21)\alpha^2 F_{\mathrm{tr}}(\omega)=\sum_{\sigma}\alpha^2 F^\sigma_{\mathrm{tr}}(\omega)=\alpha^2 F^\uparrow_{\mathrm{tr}}(\omega)+\alpha^2 F^\downarrow_{\mathrm{tr}}(\omega). \tag*{(21)}

By introducing Eq. (4) into Eq. (1), we have that ρLOVA\rho^{\mathrm{LOVA}} is also additive in the spin indices σ\sigma

ρLOVA=ρ~↑+ρ~↓,(22)\rho^{\mathrm{LOVA}}=\tilde{\rho}^{\uparrow}+\tilde{\rho}^{\downarrow}, \tag*{(22)}

which is similar to a series resistor model for the 2 spin channels. Interestingly, we note that this additive property is different from ρBTE\rho^{\mathrm{BTE}} defined in Eqs. (13) and (14) of the manuscript:

(ρBTE)−1=(ρ↑)−1+(ρ↓)−1,(23)\left(\rho^{\mathrm{BTE}}\right)^{-1}=\left(\rho^{\uparrow}\right)^{-1}+\left(\rho^{\downarrow}\right)^{-1}, \tag*{(23)}

which is a parallel resistor model for the 2 spin channels. We note that ρLOVA≈ρBTE\rho^{\mathrm{LOVA}}\approx\rho^{\mathrm{BTE}} and describe the same observable while ρ↑/ρ↓\rho^{\uparrow}/\rho^{\downarrow} and ρ~↑/ρ~↓\tilde{\rho}^{\uparrow}/\tilde{\rho}^{\downarrow} do not.

The reason for this is that the spin resolved resistivity obtained with the BTE can be interpreted as the resistivity that each spin channel would experience and is derived in the manuscript. Instead, the spin resolved resistivity obtained from α2Ftrσ\alpha^2 F^\sigma_{\mathrm{tr}} leads to ρ~↑\tilde{\rho}^{\uparrow} and ρ~↓\tilde{\rho}^{\downarrow} which are relative contributions to the total resistivity of each spin channel.

This difference can be traced back to the derivation of Ziman [1] where the resistivity is connected to the change of entropy S˙\dot{S} due to an electric current J\mathbf{J} at unit electric field E\mathbf{E} and given S˙/(J2T)\dot{S}/(\mathbf{J}^2T). Under this interpretation, it is natural to find an additive definition of resistivity with respect to the spin channels, since the entropy of the system is an extensive quantity and therefore additive (S˙↑+S˙↓)/((J↑+J↓)2T)(\dot{S}^{\uparrow}+\dot{S}^{\downarrow})/((\mathbf{J}^{\uparrow}+\mathbf{J}^{\downarrow})^2T).

VERIFICATION OF THE ELECTRICAL RESISTIVITY IN MAGNETIC COMPOUNDS

We verify the results of the implementation of the Boltzmann transport equation (BTE) in EPW with other codes. Since no public codes provide the BTE for magnetic materials, we extend the capability of the ElectronPhonon.jl package [3, 4] for the purpose of this work. We compare the resistivity of Fe and Ni on a 112×112×112112\times112\times112 k- and q-point grids in Figure S1 between EPW (and ElectronPhonon.jl, finding perfect agreement (largest difference is 2%). We note that the ElectronPhonon.jl package relies on the real-space electron-phonon matrix elements provided by EPW, thereby providing only a verification of the interpolation part of the calculations. We also report that the ElectronPhonon.jl code is about three times faster due to an inversion of the electron and phonon momentum grids loops. However, ElectronPhonon.jl currently uses Julia’s native multithreading to parallelize instead of the MPI approach employed in EPW and is therefore limited to a single compute node.

Electrical resistivity as a function of temperature for ferromagnetic iron and nickel

Figure S1. Electrical resistivity as a function of temperature for ferromagnetic iron (a) and nickel (b) computed with LDA. Green solid lines corresponds to EPW calculations while purple dashed lines corresponds to ElectronPhonon.jl.

Recovering the non-magnetic approximation in Pb

We compute the electronic dispersion and the deformation potential of Pb using LDA and LSDA. Since Pb is a Pauli paramagnet, the spin density for the spin up and spin down channels will be equivalent at every point in space:

ρ↑(r)=ρ↓(r), ∀ r∈R3.(24)\rho^{\uparrow}(\mathbf{r}) = \rho^{\downarrow}(\mathbf{r}),\ \forall\ \mathbf{r} \in\mathbb{R}^{3}. \tag*{(24)}

This implies that all interpolated quantities for the two spin channels must be identical up to numerical precision when the same set of parameters such as kk grids and cutoff energy are used for the calculation with LDA and LSDA. We confirm this in Figure S2 for the electronic band structure by comparing non-magnetic LDA with collinear magnetic LSDA. We find that the electronic band struc-

Electronic band structure of Pb with LDA and LSDA

Figure S2. Electronic band structure of Pb with (a) LDA computed with QE and (b) LDSA interpolated with EPW.

tures produce the same results for both spin channels (machine precision) and are close to the non-magnetic LDA calculation (largest difference below 0.02%). These results show that the interpolation for both spin channels works as expected and the non-magnetic solution is recovered. In addition, we compare the interpolated electron-phonon coupling gmnν(k,q)g_{mn\nu}(\mathbf{k},\mathbf{q}) in Figure S3, obtaining excellent agreement (largest difference below 0.03%). Finally, we compute in Figure S4 the resistivity of Pb with the BTE with LDA and LSDA, also obtaining a good agreement (largest difference below 0.04%).

Absolute value of the electron phonon matrix elements

Figure S3. Absolute value of the electron phonon matrix elements for bands m=n=8m=n=8 with LDA (non magnetic) and LSDA (up and down channels), compared to the non-magnetic (NM) solution.

Computed resistivity of FCC Pb

Figure S4. Computed resistivity of FCC Pb using LDA (purple) and LSDA (green).

Verification of QE and ABINIT for ferromagnetic Fe and Ni

We report the lattice parameters and the total magnetization computed with Quantum Espresso and ABINIT in Table S1.

CodeCompoundaa (Bohr)Magnetic orderMTotM^{\mathrm{Tot}} (μB\mu_{\mathrm{B}})FunctionalEcutE^{\mathrm{cut}} (Ry)k-meshPseudopotential
QEFe5.21NM0PBE10024324^{3}PseudoDojo v0.5 stand.
5.36FM - collinear2.26PBE10024324^{3}PseudoDojo v0.5 stand.
5.09NM0LDA10024324^{3}PseudoDojo v0.5 stand.
5.20FM - collinear2.05LDA10024324^{3}PseudoDojo v0.5 stand.
Abinit5.21NM0PBE10024324^{3}PseudoDojo v0.5 stand.
5.36FM - collinear2.25PBE10024324^{3}PseudoDojo v0.5 stand.
5.09NM0LDA10024324^{3}PseudoDojo v0.5 stand.
5.20FM - collinear2.04LDA10024324^{3}PseudoDojo v0.5 stand.
Experiment5.42 [5]FM2.22 [6, 7, 8, 9]----
QENi6.64NM0PBE10024324^{3}PseudoDojo v0.5 stand.
6.65FM - collinear0.65PBE10024324^{3}PseudoDojo v0.5 stand.
6.46NM0LDA10024324^{3}PseudoDojo v0.5 stand.
6.47FM - collinear0.57LDA10024324^{3}PseudoDojo v0.5 stand.
Abinit6.64NM0PBE10024324^{3}PseudoDojo v0.5 stand.
6.65FM - collinear0.67PBE10024324^{3}PseudoDojo v0.5 stand.
6.46NM0LDA10024324^{3}PseudoDojo v0.5 stand.
6.47FM - collinear0.57LDA10024324^{3}PseudoDojo v0.5 stand.
Experiment6.66 [5]FM0.62 [9]----

Table S1. Comparison between Quantum ESPRESSO (QE) and Abinit for the lattice parameter aa and the total magnetization MTotM^{\mathrm{Tot}} as well as computational parameters including planewave energy cutoff EcutE^{\mathrm{cut}}, k-point grid and pseudopotential from PseudoDojo [10].

We find excellent agreement between the two codes with a difference between the total energy below 10−510^{-5} Ry, similar to previous studies without magnetism [11, 12]. Such agreement is obtained by carefully using the same pseudopotential, FFT grid, and cutoff radius for the non-local part of the potential. We also compare the electronic bandstructure and phonon dispersion in Figure S5 and find differences as small as previously reported with non-magnetic compounds [11].

Electronic band structure and phonon dispersion of ferromagnetic Fe

Figure S5. Electronic band structure (a) and phonon dispersion (b) of ferromagnetic Fe with Quantum ESPRESSO (QE) and Abinit.

Verification of the Electron-Phonon Matrix Element Between Quantum ESPRESSO and Abinit

We compare in Figure S6 the absolute value of the directly calculated electron-phonon matrix element between Quantum ESPRESSO and Abinit of ferromagnetic Ni at k=Γk = \Gamma for m=n=5m = n = 5, which is one of the bands crossing the Fermi level. The results show good agreement, with the largest difference being 0.01 meV/bohr.

Direct evaluation of the deformation potential of ferromagnetic Ni

Figure S6. Direct evaluation of the deformation potential of ferromagnetic Ni for (a) spin up and (b) spin down at k=Γk = \Gamma for m=n=5m = n = 5, along a qq momentum line for all phonon modes ν\nu. Red dots are computed with Quantum ESPRESSO (QE) while blue squares are computed with Abinit. LDA functional was employed.

Additionally, we also compare the interpolated electron phonon matrix elements between Abinit and EPW using the same 8×8×88 \times8 \times8 q grids for the phonon part and for the Wannier interpolation of the electronic structure in EPW we use an 8×8×88 \times8 \times8 k grid. The results are shown in Figure S7 and Figure S7 for Fe and Ni where we can find a similar behavior of the interpolation methods of EPW and Abinit, although not completely equivalent close to the Γ\Gamma point.

Electron phonon matrix element for ferromagnetic Fe and Ni

Figure S7. Electron phonon matrix element for (a) Ferromagnetic Fe, and (b) Ferromagnetic Ni at k=Γ\mathbf{k} = \Gamma for m=n=5m = n = 5, along a q\mathbf{q} momentum line for all phonon modes ν\nu. Red dots are computed directly with Quantum ESPRESSO (QE) while the solid/dashed lines are interpolated with EPW/ABINIT. LDA functional was employed.

Interestingly, we note that the spin up and down electron-phonon matrix elements are visibly different for the case of iron but close in the case of nickel.

Real Space Decay of Interpolated Quantities

For an accurate interpolation, the Hamiltonian, dynamical matrices and electron-phonon matrix elements should be localized in real-space. We show in Figure S8 the decay of these quantities in real space and find that both spin channels decay in a similar way.

Real space decay of interpolated quantities

Figure S8. Real space decay of the Hamiltonian (a), dynamical matrix (b), velocity matrix element (c), dipole matrix element (d), and electron phonon matrix element in the limiting cases Re=0\mathbf{R}_{e}=0 (e) and Rp=0\mathbf{R}_{p}=0 (f).

We present the Hamiltonian in Figure S8, the dynamical matrix in (b), the velocity matrix elements in (c), the dipole matrix elements in (d) and the electron phonon matrix elements in real space g(Re,Rp)g(\mathbf{R}_{e},\mathbf{R}_{p}) for the limiting cases of Re=0\mathbf{R}_{e}=0 and Rp\mathbf{R}_{p} in (e) and (f) respectively.

In all cases, we find a decay of at least three orders of magnitude, confirming the quality of the interpolation.

Electron Phonon Coupling Strength and Superconductivity

Phonon-mediated superconductivity has been studied with different formalisms that often do not consider magnetism. In particular, the critical temperature of a superconductor can be computed using the Allen-Dynes formula [13]:

TcAD=ℏωlog⁡1.2kBexp⁡(−1.04(1+λ)λ−μ∗(1+0.62λ)),(25)T_{\mathrm{c}}^{\mathrm{AD}}=\frac{\hbar\omega_{\log}}{1.2k_{\mathrm{B}}}\exp\left(\frac{-1.04(1+\lambda)}{\lambda-\mu^{*}(1+0.62\lambda)}\right), \tag*{(25)}

where λ\lambda is the electron phonon coupling strength, kBk_{\mathrm{B}} is the Boltzmann constant, μ∗\mu^{*} is the semi-empirical Coulomb potential, and ωlog⁡\omega_{\log} is the logarithmic average of the phonon frequency (geometric average) defined as

ωlog⁡≡exp⁡(1λ∫dωωα2F(ω)log⁡ω),(26)\omega_{\log}\equiv\exp\left(\frac{1}{\lambda}\int\frac{\mathrm{d}\omega}{\omega}\alpha^{2}F(\omega)\log\omega\right), \tag*{(26)}

with α2F(ω)\alpha^{2}F(\omega) being the Eliashberg spectral function. Although this formula is generally used for the ss-wave pairing mechanism with time reversal symmetry, a BCS type formula can be obtained for the other gap channels pp, dd, or ff with spin singlet and triplet [14, 15] which requires form factors and different semi-empirical numerical parameters.

However, some authors have used the standard Allen-Dynes formula in the context of phonon mediated superconductivity in magnetic systems [16], where λ\lambda is identified with the average mass enhancement parameter. Interestingly, the use of the standard formula is valid when considering an ss-wave singlet in a commensurate collinear anti-ferromagnet (AFM) with inversion symmetry and spin degeneracy on the bands such that εk↑=εk↓\varepsilon_{\mathbf{k}}^{\uparrow}=\varepsilon_{\mathbf{k}}^{\downarrow} and gmnν↑†(k,q)=gmnν↑(k,q)g_{mn\nu}^{\uparrow\dagger}(\mathbf{k},\mathbf{q})=g_{mn\nu}^{\uparrow}(\mathbf{k},\mathbf{q}) as done in Ref. [16].

Another valid use of the standard Allen-Dynes formula is for ss-wave triplet (Sz=±1S_{z}=\pm1) with interband pairing involving transitions n,k→m,k+qn,\mathbf{k}\to m,\mathbf{k}+\mathbf{q} and m,−k→n,−k−qm,-\mathbf{k}\to n,-\mathbf{k}-\mathbf{q}. However, we must note that the λ\lambda for the pairing and the mass enhancement are not necessarily equivalent. While the mass enhancement is proportional to λmass∝∣gmnνσ(k,q)∣2\lambda_{\mathrm{mass}}\propto\lvert g_{mn\nu}^{\sigma}(\mathbf{k},\mathbf{q})\rvert^{2}, the pairing λ\lambda would be proportional to λpair∝gmnνσ(k,q)gmnνσ(−k,−q)\lambda_{\mathrm{pair}}\propto g_{mn\nu}^{\sigma}(\mathbf{k},\mathbf{q})g_{mn\nu}^{\sigma}(-\mathbf{k},-\mathbf{q}). In the case of the ss-wave pairing, we can consider that only the real part of the matrix elements are involved, λpair∝Re⁡[gmnνσ(k,q)gmnνσ(−k,−q)]\lambda_{\mathrm{pair}}\propto\operatorname{Re}[g_{mn\nu}^{\sigma}(\mathbf{k},\mathbf{q})g_{mn\nu}^{\sigma}(-\mathbf{k},-\mathbf{q})]. Under this constrains, it is interesting to consider the following inequality

2Re⁡[gmnνσ(k,q)gmnνσ(−k,−q)]≤∣gmnνσ(k,q)∣2+∣gmnνσ(−k,−q)∣2.(27)2\operatorname{Re}[g_{mn\nu}^{\sigma}(\mathbf{k},\mathbf{q})g_{mn\nu}^{\sigma}(-\mathbf{k},-\mathbf{q})]\leq\lvert g_{mn\nu}^{\sigma}(\mathbf{k},\mathbf{q})\rvert^{2}+\lvert g_{mn\nu}^{\sigma}(-\mathbf{k},-\mathbf{q})\rvert^{2}. \tag*{(27)}

If we now divide by ωqν\omega_{\mathbf{q}\nu} and N(εF)N(\varepsilon_{\mathrm{F}}), and sum over all k\mathbf{k} and q\mathbf{q} such that they belong to the Fermi surface pairs (F.S._{\mathrm{pair}}) k\mathbf{k} and −k-\mathbf{k} as well as k+q\mathbf{k}+\mathbf{q} and −k−q-\mathbf{k}-\mathbf{q}, and sum over bands mm, nn, and ν\nu. We obtain:

∑mnν∑k,q∈F.S.pair2Re⁡[gmnνσ(k,q)gnmνσ(−k,−q)]ωqνN(εF)≤∑mnν∑k,q∈F.S.pair∣gmnνσ(k,q)∣2+∣gnmνσ(−k,−q)∣2ωqνN(εF)≤∑mnν(∑k,q∈F.S.∣gmnνσ(k,q)∣2ωqνN(εF)+∑−k,−q∈F.S.∣gnmνσ(−k,−q)∣2ωqνN(εF))⟺2λpairσ≤2λmassσ,(28)\sum_{mn\nu}\sum_{\mathbf{k},\mathbf{q}\in\mathrm{F.S.}^{\mathrm{pair}}}\frac{2\operatorname{Re}[g^{\sigma}_{mn\nu}(\mathbf{k},\mathbf{q})g^{\sigma}_{nm\nu}(-\mathbf{k},-\mathbf{q})]}{\omega_{\mathbf{q}\nu}N(\varepsilon_{\mathrm{F}})} \leq\sum_{mn\nu}\sum_{\mathbf{k},\mathbf{q}\in\mathrm{F.S.}^{\mathrm{pair}}}\frac{|g^{\sigma}_{mn\nu}(\mathbf{k},\mathbf{q})|^2+|g^{\sigma}_{nm\nu}(-\mathbf{k},-\mathbf{q})|^2}{\omega_{\mathbf{q}\nu}N(\varepsilon_{\mathrm{F}})} \leq\sum_{mn\nu}\left(\sum_{\mathbf{k},\mathbf{q}\in\mathrm{F.S.}}\frac{|g^{\sigma}_{mn\nu}(\mathbf{k},\mathbf{q})|^2}{\omega_{\mathbf{q}\nu}N(\varepsilon_{\mathrm{F}})}+\sum_{-\mathbf{k},-\mathbf{q}\in\mathrm{F.S.}}\frac{|g^{\sigma}_{nm\nu}(-\mathbf{k},-\mathbf{q})|^2}{\omega_{\mathbf{q}\nu}N(\varepsilon_{\mathrm{F}})}\right)\Longleftrightarrow2\lambda^{\sigma}_{\mathrm{pair}}\leq2\lambda^{\sigma}_{\mathrm{mass}}, \tag*{(28)}

where in the last term we have used the fact that ∑k∈F.S.pair≤∑k∈F.S.\sum_{\mathbf{k}\in\mathrm{F.S.}^{\mathrm{pair}}}\leq\sum_{\mathbf{k}\in\mathrm{F.S.}} where F.S. indicates the whole Fermi surface. This result shows that λmassσ≥λpairσ\lambda^{\sigma}_{\mathrm{mass}}\geq\lambda^{\sigma}_{\mathrm{pair}} for this type of interband pairing. We note that the equality λmassσ=λpairσ\lambda^{\sigma}_{\mathrm{mass}}=\lambda^{\sigma}_{\mathrm{pair}} would only hold for systems with time reversal symmetry. Thus, we can conclude that the mass enhancement can give us an upper bound for this type of ss-wave triplet pairing where time reversal symmetry is not present. The ferromagnetic (FM) metals that we study here fall into this category and the resulting critical transition temperature should be seen as an upper bound.

We present in Table S2 the results for the FM and non-magnetic (NM) cases for both iron and nickel. As discussed in the manuscript, the electron-phonon coupling constant λ\lambda presents a pronounced difference between the two spin channels in Fe and Ni when considering the FM phase. If we compare the NM phase with the FM phase, we find that for the case of Ni FCC the value of λ\lambda is not much affected by magnetic order and only the relative importance of each spin channel is changed. This suggests a predominant role of Fermi surface nesting and the number of available states close to the Fermi level. On the contrary, for the case of Fe, the values are different with a large λ≈1.9\lambda\approx1.9 in the NM case. This is due to the low frequency phonons obtained in the NM case, further highlighting the dangers of using phonon dispersions with unstable modes. In what concerns the characteristic frequency of the system ωlog⁡\omega_{\log}, the values are rather similar between the 4 cases presented in Table S2 with ωlog⁡≈300−250\omega_{\log}\approx300-250K with the exception of NM Fe

CompoundMagnetismλ\lambdaλ↑\lambda^{\uparrow}λ↓\lambda^{\downarrow}ωlog⁡\omega_{\log} [K]TcADT_{\mathrm{c}}^{\mathrm{AD}} [K]
Fe (BCC)FM0.29610.21800.07813110.02
Ni (FCC)FM0.27670.00400.27272910.01
Ni (FCC)NM0.26060.13030.13032590.00
Fe (BCC)NM1.91240.95620.9562283.64

Table S2. Total and spin-resolved electron phonon coupling strength λ\lambda for ferromagnetic (FM) and non magnetic (NM) Fe and Ni. The characteristic frequency ωlog⁡\omega_{\log} and Allen-Dynes critical temperature TcT_{\mathrm{c}} are obtained with μ∗=0.13\mu^{*}=0.13.

BCC which is an order of magnitude smaller due to the mentioned near zero frequency modes.

At this point, one can take this characteristic frequency in the Allen-Dynes formula and the compute λ\lambda to estimate the TcT_{\mathrm{c}} of any hypothetical ss-wave superconductor. In doing so, we need to provide a reasonable value for the semi-empirical Coulomb potential μ∗\mu^{*}. Typical values range between 0.1 and 0.2 in the literature [17, 18, 19] and we choose μ∗=0.13\mu^{*}=0.13 in Table S2 for illustrative purposes, finding vanishingly small values for all cases, except NM Fe which provides a TcT_{\mathrm{c}} of nearly 4K due to the artificially large λ\lambda. For the other three cases, it is interesting to find that a phonon mediated mechanism seems rather unfavorable. These results are interesting since neither Ni nor Fe have a superconducting transition under ambient conditions and only Fe may show a superconducting transition at very low temperature Tc≤2T_{\mathrm{c}}\leq2 K under high pressure [20, 21]. The most common argument to explain this situation is that the long range magnetic order prevents it. However, the results given in Table S2 may suggest that the absence of superconductivity on these compounds might be more intrinsic rather than being suppressed by the internal magnetic field.

Momentum convergence of the resistivity

We first converge the coarse k/q grids using a fixed fine interpolated 30 × 30 × 30 grid and show the results in Figure S9 and (Figure S9) for Fe and Ni, respectively.

Resistivity using LDA with different k/q coarse and fine grids for ferromagnetic Fe and Ni

Figure S9. Resistivity using LDA with different k/q coarse and fine grids for ferromagnetic Fe (a, c), and for ferromagnetic Ni (b, d).

find that a 8 × 8 × 8 coarse k/q grid converges below 5% and offers good trade-off between computational cost and precision for both materials. With these fixed coarse grids, we repeat the process but with the fine mesh and report the results in Figure S9(c) and (d) for Fe and Ni, respectively. A convergence below 5% at 50 K and below 1% above 150 K is obtained with a fine mesh of 112 × 112 × 112 k/q mesh for both compounds. We note that 64×64×6464 \times64 \times64 grids are already accurate and provide a maximum deviation of 0.2 μΩ⋅cm0.2\ \mu\Omega\cdot\mathrm{cm}. When not explicitly reported, all resistivity results are obtained with a 8×8×88 \times8 \times8 coarse k/q grid and a fine 112×112×112112 \times112 \times112 k/q grid.

Electron band structure with PBE

We compare the electronic band structure of Ni and Fe for LDA and PBE functionals in Figure S10. We find that the exchange splitting between the two channels is enhanced when using PBE instead of LDA. The results shown in Figure S10 for both compounds show a change of the band velocity in both cases when using PBE with respect to LDA.

Electronic band structure of ferromagnetic Ni and Fe using LDA and PBE

Figure S10. Electronic band structure using Quantum Espresso for ferromagnetic Ni (a) and (b) with LDA and PBE functionals respectively, and for ferromagnetic Fe (c) and (d) with LDA and PBE functionals, respectively. Red solid lines correspond to spin up or spin majority channel. Blue solid lines correspond to spin down or spin minority channel.

Ferromagnetic Co HCP

Indicatively, we also compute the electronic and phonon band structure of ferromagnetic Co hcp with the PBE functional showing the results in Figure S11. For the electronic bandstructure, a 24×24×1224 \times24 \times12 k grid is used, and for the phonons a 8×8×48 \times8 \times4 q grid is used. Furthermore, we compute the electrical resistivity with an 8×8×48 \times8 \times4 coarse and a 96×96×4896 \times96 \times48 fine k/q grids and compare it with the experiments in Figure S12. We find that up to 300 K, the electron-phonon interactions can account for up to 50% of the experimental resistivity placing Co in between the behavior of Fe and Ni discussed in the manuscript.

Dispersion relation for electrons and phonons in ferromagnetic Co hcp

Figure S11. Dispersion relation for (a) electrons and (b) phonons in ferromagnetic Co hcp.

Figure S12. Electrical resistivity of ferromagnetic Co. Solid lines are the calculated resistivity using PBE. Black symbols represent the experimental resistivity from Refs. [22, 23].

Non-magnetic solution for Fe and Ni

Finally, we also show in Figure S13 the electrical resistivity for non-magnetic Fe and Ni using LDA and compared with experimental results. When the phonons are soft (negative), we neglect their contribution to the resistivity. In the case of Fe in Figure S13, we can clearly see a generalized overestimation of the resistivity similar to the case with a PBE functional. This behavior clearly shows the need to include magnetism in Fe. In the case

Electrical resistivity of non-magnetic Fe and Ni

Figure S13. Electrical resistivity of non-magnetic Fe (a), and Ni (b). Solid lines are the calculated resistivity using LDA Black symbols represent the experimental resistivity from Refs. [22, 23, 24].

of Ni in Figure S13, we obtain a resistivity curve that below 300 K slightly overestimates the resistivity and then above 300 K it starts to underestimate it. Although one may be tempted to think that this is something satisfactory, since below 300 K the resistivity curve is quite close to the experimental values, one should note that both the electronic dispersion and the phonon dispersion change significantly between LDA and PBE. In particular, the phonon dispersion of LDA clearly overestimates the phonon frequencies in contrast with the PBE calculation which yields a rather good match with the experiments. Thus, this better agreement might be due to error cancelations.

References

  1. [1]J. M. Ziman, Electrons and Phonons: The Theory of Transport Phenomena in Solids (Oxford University Press, Oxford, 1960).DOI
  2. [2]G. Grimvall, The electron-phonon interaction in metals (North-Holland Pub. Co.: Sole distributors for the U.S.A. and Canada, Elsevier North-Holland, 1981).
  3. [3]J.-M. Lihm and C.-H. Park, Nonlinear Hall effect from long-lived valley-polarizing relaxons, Phys. Rev. Lett. 132, 106402 (2024).DOI
  4. [4]J.-M. Lihm, Electron Phonon.jl (2025).
  5. [5]N. M. Rosengaard and B. Johansson, Finite-temperature study of itinerant ferromagnetism in fe, co, and ni, Phys. Rev. B 55, 14975 (1997).DOI
  6. [6]J. F. van Acker, Z. M. Stadnik, J. C. Fuggle, H. J. W. M. Hoekstra, K. H. J. Buschow, and G. Stroink, Magnetic moments and x-ray photoelectron spectroscopy splittings in fe 3s core levels of materials containing fe, Phys. Rev. B 37, 6827 (1988).pubmed.ncbi.nlm.nih.gov/9943953
  7. [7]V. Pasyuk, O. Mc Grath, H. Lauter, A. Petrenko, A. Liénard, and D. Givord, Ground state moment reduction in an ultra-thin W (110)/Fe (110)/W (110) film, Journal of magnetism and magnetic materials 148, 38 (1995).
  8. [8]D. Bardos, Mean magnetic moments in bcc Fe – Co alloys, Journal of Applied Physics 40, 1371 (1969).DOI
  9. [9]H. Danan, A. Herr, and A. Meyer, New determinations of the saturation magnetization of nickel and iron, Journal of Applied Physics 39, 669 (1968).DOI
  10. [10]M. van Setten, M. Giantomassi, E. Bousquet, M. Verstraete, D. Hamann, X. Gonze, and G.-M. Rignanese, The pseudodojo: Training and grading a 85 element optimized norm-conserving pseudopotential table, Computer Physics Communications 226, 39–54 (2018).DOI
  11. [11]S. Poncé, G. Antonius, P. Boulanger, E. Cannuccia, A. Marini, M. Côté, and X. Gonze, Verification of first-principles codes: Comparison of total energies, phonon frequencies, electron–phonon coupling and zero-point motion correction to the gap between abinit and qe/yambo, Computational materials science 83, 341 (2014).DOI
  12. [12]S. Poncé, J.-M. Lihm, and C.-H. Park, Verification and validation of zero-point electron-phonon renormalization of the bandgap, mass enhancement, and spectral functions, npj Computational Materials 11, 117 (2025).DOI
  13. [13]P. B. Allen and R. C. Dynes, Transition temperature of strong-coupled superconductors reanalyzed, Physical Review B 12, 905–922 (1975).DOI
  14. [14]P. W. Anderson and P. Morel, Generalized bardeen-cooper-schrieffer states and the proposed low-temperature phase of liquid he3, Phys. Rev. 123, 1911 (1961).
  15. [15]M. Houzet, Applications of symmetries in superconductivity, in EPJ Web of Conferences, Vol. 22 (EDP Sciences, 2012) p. 00014.DOI
  16. [16]R. Zhang, Y. Wang, M. Engel, C. Lane, H. Miranda, L. Hou, S. Chowdhury, B. Singh, B. Barbllini, J.-X. Zhu, R. S. Markiewicz, E. K. U. Gross, G. Kresse, A. Bansil, and J. Sun, Magnetism-enhanced strong electron-phonon coupling in infinite-layer nickelates, Phys. Rev. B 112, L241115 (2025).DOI
  17. [17]M. Kawamura, Y. Hizume, and T. Ozaki, Benchmark of density functional theory for superconductors in elemental materials, Phys. Rev. B 101, 134511 (2020).DOI
  18. [18]B. Lilia, R. Hennig, P. Hirschfeld, G. Profeta, A. Sanna, E. Zurek, W. E. Pickett, M. Amsler, R. Dias, M. I. Eremets, et al., The 2021 room-temperature superconductivity roadmap, Journal of Physics: Condensed Matter 34, 183002 (2022).
  19. [19]M. Bercx, S. Poncé, Y. Zhang, G. Trezza, A. G. Ghez eljehmidan, L. Bastonero, J. Qiao, F. O. von Rohr, G. Pizzi, E. Chavazzo, and N. Marzari, Charting the landscape of bardeen-cooper-schrieffer superconductors in experimentally known compounds, PRX Energy 4, 033012 (2025).DOI
  20. [20]K. Shimizu, T. Kimura, S. Furomoto, K. Takeda, K. Kontani, Y. Onuki, and K. Amaya, Superconductivity in the non-magnetic state of iron under pressure, Nature 412, 316 (2001).pubmed.ncbi.nlm.nih.gov/11460158
  21. [21]D. Jaccard, A. Holmes, G. Behr, Y. Inada, and Y. Onuki, Superconductivity of ε-Fe: complete resistive transition, Physics Letters A 299, 282 (2002).DOI
  22. [22]G. K. White and S. B. Woods, Philosophical Transactions of the Royal Society of London. Series A, Mathematical and Physical Sciences 251, 273–302 (1959).
  23. [23]M. J. Laubitz, T. Matsumura, and P. J. Kelly, Transport properties of the ferromagnetic metals ni· nickel, Canadian Journal of Physics 54, 92–102 (1976).
  24. [24]W. Fulkerson, J. P. Moore, and D. L. McElroy, Comparison of the thermal conductivity, electrical resistivity, and seebeck coefficient of a high-purity iron and an armco iron to 1000°c, Journal of Applied Physics 37, 2639–2653 (1966).DOI

Paper details

vlnc:W_ybvdyz7d
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents